The trend was already tracing itself through the corridors of power, but the world’s most powerful defence department has fallen, most heavily, for the use of artificial intelligence at virtually every level of its operations. Even before the Pentagon’s intoxicating pursuit of artificial intelligence was announced, the Defence Advanced Research Projects Agency (DARPA) was already running simulated aerial trials of combat engagements between AI agents and seasoned pilots.
In December 2025, the US Department of Defense (DOD) revealed plans for a rapid and extensive release of the generative AI GenAI.mil platform, one utilising advanced models intended for use for over 3 million military personnel, civilian employees and contractors. Google Cloud’s Gemini or Government products was the first off the rank in a described “one-stop shop,” a generous effort on the part of bureaucracy to hail the machine and herald the start of full-throated cognitive surrender.
In his December email to employees in a department the Trump administration prefers to call the War Department, Secretary of Defense (War) Pete Hegseth announced that the GenAI.mil platform signalled “a giant step toward mass AI adoption across the department. This tool marks the beginning of a new era, where every member of our workforce can be more efficient and impactful.”
In his January memorandum to the various leaders of the military community, Hegseth barked about the need for the US military to “build on its lead over our adversaries in integrating this technology, established during President Trump’s first term, to make our Warfighters more lethal and efficient.” To do so would involve the Pentagon’s acceleration of US “Military AI dominance by becoming an ‘AI-first’ warfighting force across all components, from front to back.” (Why not sideways, above and below?)
Trouble, however, arose when Anthropic placed restrictions on the use of Claude and its products for conducting mass surveillance of US citizens or use in lethal autonomous systems. The Pentagon, in turn, demanded the use of any genAI products it purchased for any designated “lawful purposes.” For its stubbornness, Anthropic was designated a “supply-chain risk” by the department, its products removed from all national security assets and systems. (The company’s efforts to have the ban overturned have had mixed results.)
Notwithstanding this hiccup—a minor one in the scheme of things—the contracts between AI companies and the Pentagon make it clear that the former are meant to provide “risk forecasting, and threat ideation exercises.” “Frontier AI companies forecasting and in a better understanding of the risks that might be associated with subsequent near-frontier AI models or features, to avoid strategic surprise and to mitigate associated risks.”
The Pentagon has since added other commercial generative AI products to the GenAI.mil portal, including Starshield AI’s Grok and OpenAI’s ChatGPT Mil. In August, a source self-described as a “War Department official” told DefenseScoop that such access to “the resilient American technology stack will give warfighters the tools they need to act with confidence and safeguard the nation against any threat.”
ChatGPT Mil also serves to organise salient files and conversations that take place over projects lasting months in one place, including an “Offline Search capability” that supplies detailed citations. The facility can be refreshed over time to reflect new information. Joe Larson, former senior Pentagon official turned OpenAI vice president, very much an exponent of the defence-civilian revolving door, explained this integration as “grounded in our national security principles, pairing access to powerful tools with clear safeguards, meaningful human judgment, and accountability for high-stakes decisions.”
Grok for Government will also, according to the department, furnish personnel through GenAI.mil with “advanced capabilities, including deep-thinking inference, adaptive reasoning modes (Auto, Fast, and Expert), customizable workspaces, persistent projects and reusable ‘playbooks’ that capture and scale institutional knowledge.”
These systems presage a broader surrender of the US military establishment to the machine, its alluring fetishes, and its acclaimed shamanistic potential. The guidance on “Accelerated Mission Software” signed by Pentagon CIO Kirsten Davies on 31 August is yet another indication of this numbing trend. It notes the recognition on the part of the department “that AI-assisted software development is a significant force multiplier, for the delivery of speed quality.” Something to be said about the guidance is its acknowledgment that human agency “remains fully accountable for the security, functionality, and integrity of any code generated or modified using AI.” Any code with an AI-provenance “will be considered unverified input, and its use does not absolve the developer or the government of responsibility for the resulting work product.”
Such a cautionary note may well be temporary. In his 30 September Quantico speech, Hegseth announced the creation of the Autonomous Warfare Command “designed to scale autonomous warfare across the Joint Force as a 4-star functional combatant command by 1 October 2027.” In his memorandum on the subject, the Secretary is filled with effusive babble on the new command, which “will define joint needs and integrate Service-provided forces in support of geographic combatant commanders. It will test equipment and force designs with realistic opposition in demanding exercises and rugged environments, modify rapidly, and field what works.” Just the sorts of things to make humans feel safe.
The consequences of such AI integration and excessive reliance are proving and will prove critical. The disclosures by OpenAI and Anthropic that semi-autonomous large language models conducted coordinated attacks on networks owned by other companies and, in the case of Australia, various government bodies and affiliates, including the Australian Institute of Health and Welfare (AIHW), should engender teeth-chattering doubt among sceptics. While US President Donald Trump insists with consistent daftness that AI firms self-regulate the field, an independent assessment of the risks rather than a self-reporting format, has, at the very least, the merits of greater accountability. Heidy Khlaaf, a former safety engineer at OpenAI and currently chief scientist at the AI Now Institute, is of the view that trusting companies to consider the risks produced by their own products is not just a conflict of interest but a “subversion of democratic processes when AI labs are allowed to take over the arbitration of risk determinations with life-or-death consequences.” Then again, companies were never much for democracy.
